interesting-people message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]

Subject: [IP] more on skype

  • From: David Farber <>
  • To:
  • Date: Fri, 23 Jun 2006 13:13:32 -0400

Begin forwarded message:

From: Bob Frankston <>
Date: June 23, 2006 1:09:12 PM EDT
Cc:, &quot;'Christian de Larrinaga'&quot; <>
Subject: RE: [IP] skype

This makes me think of something Butler Lampson said to me in the 60&#x2019;s &#x2013; deal with the extreme cases and the middle will take care of itself. Or something like that &#x2013; the point here is that if you can program around trouble well enough why not make that the norm rather than a special case? You can then choose the extreme of depending upon reliable circuits or total independence. What&#x2019;s the advantage of depending on a network promising to be E2E when you can&#x2019;t really trust the claim anyway?

The importance of Skype is that it&#x2019;s the real face of end-to-end. It may seem like a paradox but to get true end-to-end behavior you cannot depend on the underlying transport being end-to-end. If you&#x2019;re at a conference and you just connect to the Ethernet jack you na&#xEF;vely assume you&#x2019;ve got E2E but you don&#x2019;t know anything about all segments of the network and it is likely that someone will be watching your packets, even if only for entertainment purposes.

Just as you can create a reliable path over an unreliable network using a protocol like TCP you can take a compromised network and get end-to-end behavior by encrypting and not being dependent upon the accidental addressing structure. In both cases there are limits to how well you can do it but it&#x2019;s important to observe that you can and that you must.

Na&#xEF;vely assuming that the underlying network is truly E2E is akin to being at the mercy of an underlying reliable circuit architecture. You pay a price for favors you don&#x2019;t want while putting yourself at risk.

The real power of Skype is that you don&#x2019;t need a single underlying network and can composite a path through an arbitrary infrastructure. This has similarities to UUCP routing which didn&#x2019;t depend on the existence of any network &#x2013; just a path through cooperating systems.

If you depend on &#x201C;firewalls&#x201D; you are very far from E2E &#x2013; it seems perverse to argue that you should depend a network being E2E in order to preserve the non E2E aspects.

Skype relies upon a unified directory system but a fully distributed version represents the future of the Internet. It wouldn&#x2019;t depend on &#x201C;Internet Inc&#x201D; and would come far closer to the idea of routing around problems. To those dependent upon the governed hierarchical Internet a more distributed approach may seem alien. But then today&#x2019;s Internet is alien to the Bellista.

The successor architecture would be unmappable &#x2013; you will only be found if you choose to be. Today you are findable by default and all sort of interesting stuff comes bubbling to the surface and, perhaps, we are all-the-better for it. But future Googles will have of to entice people to have their innermost contents listed or find creative ways to ferret out the information.

-----Original Message-----
From: David Farber []
Sent: Friday, June 23, 2006 08:37
Subject: [IP] skype

Begin forwarded message: forwarded message:

From:   Christian de Larrinaga <>

de Larrinaga <>

Date:    June 23, 2006 5:05:27 AM EDT

To: David Farber <>,

Subject:            Re: [IP] Digest 1.1054 for ip

Skype uses the &#x201C;accidental&#x201D; topology of the network as the platform for its addressing. How else does it do it? But because Skype assumes its users are likely not to have static IP addresses Skype has a way to route to these users. Its P2P meshing is not a consequence of this but a necessary party to making this possible.

There is also a potential cost to this as anybody who uses Skype through a firewall knows as you open up undefined ports to unknown hosts.

It would be so much easier to establish application level addressing schema&#x2019;s including for Skype type services over an end to end underlying addressing architecture. Being able to program one&#x2019;s way out of trouble is clever but it is intelligent not to have to.

Christian de Larrinaga

Bob Frankston <>wrote.

On 23 Jun 2006, at 01:28, wrote:

You don&#x2019;t want static IP addresses &#x2013; that&#x2019;s like putting a concrete

base on your cordless phone. Look at Skype as an example &#x2013; it creates

its own stable addresses without tying them to the accidental

topology of the network. You get P2P meshing as an accidental


Where I do agree is that local traffic tends to be local and peering

is a business relic not a technology.

You are subscribed as

To manage your subscription, go to

Archives at: people/

You are subscribed as
To manage your subscription, go to

Archives at:

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]

Search: Match: Sort by:
Words: | Help


Powered by eList eXpress LLC